School of Economics and Management

Transcription

1 LUND UNIVERSITY School of Economics and Management Firm Level Factors Affecting Liquidity The Swedish Stock Market Authors: Jose Roberto Pagoada Kim Jonsson MASTER THESIS Supervisor: Birger Nilsson Date: 3 rd, June 2015 i

2 FIRM LEVEL FACTORS AFFECTING LIQUIDITY- THE SWEDISH STOCK MARKET By Jose Roberto Pagoada and Kim Jonsson A thesis submitted in partial fulfillment of the requirements for the degree of Master of Science in Finance LUND UNIVERSITY 2015 Approved by Chairperson of Supervisory Committee Program Authorized to Offer Degree Date ii

3 Abstract Title: Firm Level Factors affecting Liquidity-The Swedish Stock Market Authors: Jose Roberto Pagoada and Kim Jonsson Advisor: Birger Nilsson Keywords: Liquidity, Illiquidity, Amihud measure, Bid Ask Spread, Turnover, Liquidity Premium, LCAPM Purpose: The purpose of this study is to further look at which firm-adjustable factors affect the liquidity of a firm's stock in the Swedish stock market. We also want to determine which out of these suggested factors have a significant effect on liquidity across different proxies of liquidity. Finally we aim to determine whether our findings are consistent with previous research findings. Methodology: A quantitative approach with the interpretation of the results from panel data regressions. Theoretical framework: Liquidity Capital Asset Pricing Model (LCAPM) Empirical foundation: A sample of 433 firms during the time period Conclusions: We come to the conclusion that the firm s assets liquidity and the ownership structure, more specifically the cash and cash equivalents and the free floating shares, have a positive relationship with liquidity for firms on the Swedish Stock market. These findings are consistent with other studies in other markets. Our data did however not provide significant results to establish a relationship between liquidity and capital structure or dividend payout policies. i

4 Table of Contents Abstract... i Table of Contents... ii List of Tables... iv Chapter I: Introduction... 1 Chapter II: Theoretical Background... 6 A.Liquidity... 6 A.1 Associated Costs... 7 A.1.1 Adverse Selection Costs... 8 A.1.2 Opportunity Costs... 9 A.1.3 Direct Costs... 9 A.2 Liquidity Measurements A.2.1 Bid-Ask Spread A.2.2 Amihud Measure A.2.3 Turnover B. Liquidity Premium B.1 Estimation Methods C. Liquidity Premium and the link to corporate finance D. Firm Liquidity E. Earlier Research in Sweden Chapter III: Methodology A. Sample and Data B. Liquidity Variables C. Explanatory Variables and Control Variables C.1 Explanatory Variables C.2 Control Variables D. Summary Statistics E. Regression F. Reliability and Validity F.1 Data testing Chapter IV: Results and Analysis A. Correlations B. Regression Results and Analysis Chapter V: Conclusion Appendix Appendix A ii

5 Appendix B Appendix C Appendix D References iii

6 LIST OF TABLES Table Number Page 1. Liquidity Variables-Summary Statistics Independent and Control Variables-Summary Statistics Correlation Liquidity Variables Correlation Independent and Control Variables Final Regression- Bid-Ask Spread Robustness check on Bid-Ask Spread Final Regression- Amihud Measure Robustness check on Amihud Measure Final Regression- Turnover Robustness check on Turnover Final Regression Comparison Table iv

7 I. Introduction Amihud and Mendelson first established a link between expected excess returns and liquidity in Their study arrived to the conclusion that investors require higher returns on less liquid stock. Studies such as Brennan and Subrahmanyam (1996), Brennan et al (1998) and, Pastor and Stambaugh (2003) have all validated the findings of Amihud and Mendelson (1986), indicating that the level of liquidity indeed has an effect on expected returns and as a result also on asset pricing. They distinguish between the level of liquidity and the liquidity risk where the former is the amount of liquidity measured for example by the bid-ask spread and the latter is the risk that the illiquidity level will deviate from its expected value. In a later paper Amihud & Mendelson (1991) describe their 1986 results as the liquidity effect. The liquidity effect is compared to the effect of risk on capital assets. Just like risk-averse investors require additional compensation for bearing higher risks, investors will require additional compensation for bearing additional illiquidity on its assets. The liquidity effect shows how the microstructure of security markets is relevant for firms as well as the value of financial policies targeting to increase the liquidity of individual securities and of the market as a whole. Private benefits can thus be created by increasing securities liquidity since issuers are able to sell more liquid stock at higher prices. Liquidity is therefore essential when developing a company s financial policies regarding financial instruments. Similarly financial analysts must include liquidity in their stock valuations. Financial advisors also should aim to match the investors liquidity needs with the assets included in the portfolio. 1

8 This has inspired a vast number of studies to look at variables that affect liquidity and its associated costs. Additional implications have been explored given this liquidity effect including implications for corporate management, for capital structure (O Connor, Lesmond & Senbet, 2008; Lipson & Mortal, 2009), for security design regarding the number of markets in which a stock is listed (Amihud, Lauterbach and Mendelson, 2003), for payout policies regarding dividends and buybacks (Banerjee, Gatchev, Spindt, 2007; Brockman, Howe and Mortal, 2008) as well as corporate investment decisions, corporate governance and managerial incentives (Amihud and Mendelson, 2012). Variables that affect liquidity can be separated into systematic factors (non-firm-adjustable) and firm-adjustable factors. Systematic factors concern such factors that cannot be affected by an individual firm which are due to macroeconomic factors. Firm-adjustable factors include factors that an individual firm can modify on their own. Extensive literature exists on individual factors in the form of suggestions of liquidity-enhancing policies since there is a connection between these factors, liquidity and returns separately. These include going public, increasing the small investor base, increasing the information disclosure, modifying the dividend payout policy, increasing the firm cash liquidity, avoiding fragmentation when issuing securities. Researchers still have the task to determine which of the policies have a higher impact in increasing the securities liquidity and to explore additional liquidity-enhancing policies. The purpose of this study is to further look at which firm-adjustable factors might affect the liquidity of a firm's stock in the Swedish stock market. We also want to determine which out of these suggested factors have a significant effect on liquidity across different proxies of liquidity. Finally we aim to determine whether our findings are consistent with previous research findings. This has a potential to help Swedish firms decide which action the most beneficial in terms of increasing liquidity and as result to their stock price. Our focus is on the Swedish stock market due to the previous lack of studies conducted in this setting. It is worth mentioning that the Swedish stock market differs from for example the US market in the sense that there are no 2

9 real market makers for the stocks given that trading is executed primarily electronically. For our study we therefore consider the electronic order book as the market maker. Our models will be estimated by running panel data regressions using a number of explanatory variables representing four firm-adjustable factors: a) the dividend policy measured by its dividend yield, b) the firm s assets liquidity measured by cash and cash equivalents and by the quick ratio, c) information disclosure and capital structure both proxied by a single variable debt level, and d) the firm s ownership structure measured by its percentage of free floating shares and by its percentage of institutionally held shares. Three proxies are used for measuring liquidity: the bid-ask spread, the Amihud measure and the turnover. It is important to have in mind that the costs associated with implementing the suggested financial policies are out of the scope of this study. There is as always a cost versus benefit analysis that needs to be performed by a company in order to determine if said policy effect on liquidity will outweigh the implementation costs. This is a topic that requires further research. As previous studies have shown, there is an economic and significant relation between the expected return and the liquidity of a firm's stock. This relationship affects the firm s stock price, which then can be lower than what the fundamental value indicates. This result has implications for firms because if they decrease the costs that arise from liquidity they can improve their stock price, everything else being equal. Amihud and Mendelson (2000) showed that this increase in stock price could be achieved without changing other important corporate characteristics such as expected earnings, growth, assets etc. Furthermore, an increase in liquidity would not only benefit the individual company by an increased stock price but also the financial market as a whole. Improved liquidity will result in smaller spreads and a smaller price impact. In the end, this will bring the real market closer to the underlying principle of perfectly liquid markets made in asset pricing theory eliminating liquidity as a friction and resulting in more efficient capital allocation. 3

10 Our findings reveal certain similarities with previous studies with some of the firm-adjustable factors although for some firm-adjustable factors the data does not allow to make any significant conclusions. We show how the dividend policies have a significant relationship with liquidity but no conclusion can be determined regarding the sign (positive or negative) of this relationship. The firm s assets liquidity measured by the cash and cash equivalents has a positive relationship with liquidity and is consistent with previous studies (Gopalan, Kadan and Pevzner, 2012). However when the quick ratio measurement is used we observe a significant negative relationship. A possible explanation for this result could be the fact that this measurement captures a different aspect of the asset s liquidity which causes the negative relationship. Finally, data showed that the ownership structure of a firm is significantly related to the liquidity of a firm. The number of free-floating shares is positively related to liquidity as other studies had found (Amihud et al. 1999). Based on these results firms in Sweden may align their financial policies to increase their firm s assets liquidity and increase their free floating shares in order to improve their stock s liquidity. On the other hand we obtained inconclusive results for two variables. Our results did not reveal any significant relationship between our variable (DEBT_LEV) measuring the debt level of a firm, representing both the capital structure and the information disclosure of the firm, and liquidity. Moreover the findings on the institutional shareholding were only significant for one of the proxies of liquidity (ILLIQ), thus not allowing us to make substantial conclusions regarding the relationship with liquidity. Further studies remain to be conducted to determine whether dividend payout policies are positively or negatively related to the liquidity of a stock as well the existence of a significant relationship between capital structure and information disclosure and liquidity in Sweden. The remaining four chapters of the thesis include in Chapter II a theoretical background, Chapter III a description of the methodology. Chapter IV presents the results of our study and 4

11 Chapter V has a summary of the study and concluding remarks. Chapter II s theoretical background discusses the different literature on liquidity, the associated liquidity premium and a review of the literature that establishes a connection of the micro market structure and its implications on corporate finance. It concludes with a review of prior research conducted on the Swedish stock market. Chapter III details the data used and the sources where it was obtained. It follows with a description of the liquidity variables, the independent variables and the control variables. It ends with summary statistics of the variables, the description of the models for the regression and a discussion of validity and reliability of the data. Chapter IV presents the final results of the study as well as a discussion analyzing and interpreting these results. Lastly, Chapter V summarizes all the results and presents our conclusions from the study. 5

12 II. Theoretical Background A. Liquidity The concept of liquidity is a well-researched topic, many scholars and researchers have investigated the costs that are associated with liquidity and the implications that these have for asset pricing and price efficiency. Costs that are associated with liquidity are primarily related to the notion of information and friction costs, which include adverse selection, opportunity, and direct costs such as commissions & fees. (Amihud & Mendelson, 2000; Agarwal, 2009). These costs will be covered in more detail later on in this chapter. There are two distinct types of liquidity, the first one called trading liquidity indicates how fast something can be transacted and to which roundtrip cost, and the other one called funding liquidity indicates the ease with which a firm can obtain funding (Hibbert et al, 2009; Amihud et al, 2005). The two concepts are somewhat interlinked, especially in times of economic crisis. We will however focus on trading liquidity in this paper, and the fact that an illiquid asset will take a longer time to transact than a liquid one, which in turn have implications for asset pricing (Amihud & Mendelson, 2000; Amihud et al, 2005). Much theory of asset pricing is built on some assumptions such as no arbitrage, equilibrium and utility maximization, together with the notion of perfect liquid markets without frictions. These assumptions mean that every asset can be traded at any time with no additional cost (Amihud et al, 2005; Pennacchi, 2008). Acharya and Pedersen (2005) derive a liquidity-adjusted capital asset pricing model (LCAPM) to capture the impact of liquidity risk and commonality on asset pricing. Unlike the traditional CAPM model, which assumes no additional costs when pricing the asset, Acharya and Pedersen (2005), incorporates trading costs or frictions to the model: i E t (r t+1 ) = (r f ) + E t (c t+1 i i cov t (r t+1 ) + t i c t+1 m m, r t+1 m ) var t (r t+1 c t+1 c m t+1 ) [1] 6

13 where, t =E t (r m t+1 c m t+1 r f ) is the risk premium In this version, the variables are conditional on the information available up to time t, where i m r t+1 is the expected return for the stock, r t+1 is the market return, r f is the risk-free rate and i c t+1 is the liquidity level premium for the stock. The liquidity level premium represent transactions costs, such as fees and the bid-ask spread, and in a wider setting, they can also represent search and delay costs for the particular asset (Acharya & Pedersen, 2005). This model also captures the liquidity risk premium, represented by the second term in the equation; this captures the uncertainty in the liquidity cost itself. The relationship is that total liquidity premium consists of both liquidity level premium and liquidity risk premium as captured by the LCAPM. This means that if illiquidity is zero (c m t+1 i = c t+1 = 0), there will be no liquidity frictions and the equation above equals the traditional CAPM model. This makes the liquidityadjusted capital asset pricing model consistent with earlier research that have shown that an illiquid asset can trade below its fundamental value, which is the present value of its dividends, everything else being equal. This price difference is as discussed above called the liquidity premium (Lester et al. 2012). A.1. Associated Costs The total cost associated with the liquidity level premium can be divided into information costs and real friction costs. Information cost relates to the fact that investors hold different information, implying that there is a risk that you are dealing with an informed investor when executing a trade. These costs are often referred to as adverse selection costs (Amihud & Mendelson, 2000). The notion of real friction costs relates to costs associated with the actual trade, such as brokerage fees and search costs. These types of costs are often referred to as opportunity costs and direct costs. In this section we will cover these costs in more detail. 7

14 A.1.1 Adverse selection costs The notion of adverse selection costs can also be referred to as information costs and is an important component of total liquidity costs as captured by the bid-ask spread (Amihud & Mendelson, 2000; Agarwal, 2009). Since the bid-ask spread represent the highest and lowest price at which an investor can buy and sell an asset instantaneously it represent liquidity. The spread difference in itself, which is the actual difference between the bid and ask price in absolute numbers, also indicates the cost of liquidity for limited quantities. When selling a larger quantity the investor can actually move the sell price in a negative way, which in turn also might hurt the investor because he gets a lower price. This sort of transaction usually cause the stock price to decline, but part of the effect is just temporary and part of it is a lasting effect. When talking about market impact costs we talk about the temporary effect that a large transaction have on stock price (Amihud & Mendelson, 2000). These types of costs are called adverse selection costs since the information that make the transaction happen may not be included in the market price of the stock. Brennan and Subrahmanyam (1996), Brockman et al (2009) and Aslan et al. (2011) all look at these adverse selection cost by differencing between informed and uninformed investors. Informed investors are those who have superior information about a firm and its stock value, while uninformed investors are those who trade based on liquidity. Trading on liquidity means that they buy and sell stocks when they have a shortage or abundance of money. These two groups cannot easily be separated by the market maker and as a result the bid-ask spread will be larger when they think they are dealing with informed investors (Amihud & Mendelson, 2005). Evidence shows that when the identity of a seller of a large quantity of stocks is unknown the market maker will assume that the seller is informed. This will then cause the spread to widen (Amihud & Mendelson, 2000). Other studies have validated this finding by looking at companies with a large number of shareholders and the effect that this have on the bid ask spread. Benston and Hagerman (1974) found a negative correlation between a stocks bid-ask spread and the number of shareholders. Brockman et al. (2009) also found a similar result when 8

15 looking at the effect of block ownership on the bid-ask spread. Brockman et al. (2009) define block holders as shareholders that hold more than 5% of a firms total shares, They found that block ownership tend to increase the spread and as a result decrease liquidity, and increase the adverse selection component. A.1.2 Opportunity costs Opportunity costs relate to the notion of friction costs and captures search frictions and execution delays. The buyer or a seller of a large quantity of stocks faces a choice between immediate transactions or searching for a buyer/seller that is willing to pay a better price. The immediate transaction includes turning to a market maker which buys the security at a discount, this transaction might also include the risk of market impact (Amihud & Mendelson, 2000; Amihud et al. 2005). When choosing to search for a better counterparty there is a delay/opportunity cost related to this action arising from the fact that the market might move, this cost is hard to measure but according to some studies, it can be substantial (Amihud & Mendelson 2000). A.1.3 Direct costs Direct costs relate to the notion of exogenous transaction costs, and include the cost that a buyer or seller faces every time a transaction is being carried out. This includes things such as broker fees, transaction fees, taxes etc. (Hibbert et al., 2009). When facing a transaction investors look at the total cost of buying or the net revenue from selling, so they include all types of cost in their assessments (Amihud & Mendelson 2000). Other studies by Chan and Lakonishok (1995) showed that there is a difference between block owners (owners of 5% of companies stocks) and individual investors, and their willingness to pay commission fees in return for less market impact of their trade. However, with the ease and help of the internet commission fees has declined sustainably during the past decade. 9

16 A.2 Liquidity Measurements A.2.1 Bid-Ask Spread As previously mentioned the bid-ask spread works as a proxy for liquidity because it represents the cost at which an asset can be either sold or bought instantaneously. This means that the spread itself gives an indicator of the liquidity; a wider spread indicates illiquidity. It is also possible to compute the difference between the ask price and the bid price to get an absolute number on liquidity cost for limited quantities (Amihud & Mendelson, 1986) (Amihud et al., 2005). The relative quoted bid-ask spread(qs it ) is measured as followed: QS it = [Ask it Bid it ] (Ask it +Bid it )/2, [2] where Ask it, Bid it stands for the ask and bid price for company i, at time t. The spread also depends on adverse selection costs, which arises because the market makers cannot distinguish between informed and uninformed investors (Aslan et al., 2011; Brockman et al, 2009). Furthermore, the spread should also be wide enough to cover the inventory cost (the cost of holding on to the asset before they can sell them) for the market makers (Amihud et al., 2005). These attributes together with the accessibility of the actual spread quotes in the market make it a popular proxy for liquidity in earlier research. A.2.2 Amihud Measure The Amihud measure was proposed by Yakov Amihud in 2002 in order to examine the relationship of stock returns and liquidity over time. It is defined as the daily ratio of absolute stock return over its dollar volume averaged over some period of time. It is designed to capture the price impact of trading by measuring the daily price response related to a dollar of trading. Illiquidity (ILLIQ it ) is measured as follows: 10

17 ILLIQ it = R it 100 V it, [3] where Rit is the price return on asset i, at time t, and Vit is the volume of trading. This measure has been widely used among researchers given the wide availability and easiness to obtain the data to calculate the illiquidity of an asset. (De Jong and Driessen, 2012; Han and Zhou, 2007; Dick-Nielsen et al. 2012; Acharya and Pedersen, 2005). Goyenko, Holden and Trzcinka (2009) conducted a study to compare the different measures and determine to what extent they captured liquidity. They concluded that the Amihud measure is a good proxy for capturing price impact and they recommend this measure to be used if no high-frequency data are available. A.2.3 Turnover Another measurement that is used to represent liquidity is the turnover. Brooke et al. (2000) uses turnover together with the bid-ask spread to measure liquidity. The turnover is simply the total trading volume of an asset over a specified period divided by the number of outstanding assets during the same period. Turnover (TURN it ) is measured as follows: TURN it = Number of assets traded Number of outstanding assets. [4] This measurement gives an indication of how transacted a specific asset is and studies such as Amihud and Mendelson (1986) found a negative relation between turnover and illiquidity costs. Indicating that a higher turnover is associated with higher liquidity, which is consistent with earlier findings regarding size determinants of the bid-ask spread. However, researchers such as Agarwal (2009) argue that the turnover measurement is a noisy measure for liquidity since it also captures other effects and as a result, it should be used as a complement. 11

18 B. Liquidity Premium Based on standard asset pricing theory, in a frictionless market, assets with the same cash flows will have the same price (Hibbert et. al., 2009). It has been shown however that market frictions such as transaction costs affect asset pricing. Another aspect of asset pricing has been the assumption that investors are risk averse, which has lead research to identify and measure the different risks associated with each asset (Bossaerts, Plott and Zame, 2000). Empirical research provides strong support to the theory that riskier assets require a higher rate of return from investors. In other words investors need compensation for bearing risks, which is known as equity risk premium. Liquidity as reviewed in the previous section is a very broad concept that has many aspects and applications in financial economics, asset pricing, and corporate finance. Many studies have researched the link that exists between liquidity and asset prices, hypothesizing and empirically proving that liquidity is also priced in financial assets. The existence of a liquidity premium has implications for asset pricing since it changes the underlying assumptions on which it rests (Amihud et al, 2005). The concept of liquidity premium has had a degree of complexity given that all assets have varying levels of illiquidity (Hibbert et. al., 2009). Nonetheless it has been defined as the difference in excess return, or the difference in stock price between an illiquid asset and its identical liquid asset counterpart. This has caused three main issues when determining the existence of a liquidity premium. First, researchers have had to identify and develop proxies to measure liquidity to allow for comparison. Second, different liquidity premiums exist for different asset classes (Amihud and Mendelson, 2012) such as corporate bonds (De Jong and Driessen, 2012; Han and Zhou, 2007), equity (Acharya and Pedersen, 2005; Bekaert et al., 2007), government bonds (Amihud and Mendelson, 1991; Longstaff et al., 2005), covered bonds (Breger and Stovel, 2004). And lastly there have been many different methods to model and measure liquidity premium, which will be discussed next. 12

19 One of the first studies to look at the implications of liquidity on asset pricing was Amihud & Mendelson (1986), which looked at asset pricing in relation to the bid-ask spread. The reason for using the bid-ask spread is that it makes a simple proxy for illiquidity since the ask price includes a premium for immediate buy, and the bid price include a discount for immediate sale. In other words, a larger spread indicates a more illiquid stock and as a result, an investor should demand a higher expected return to compensate for this everything else being equal. Amihud & Mendelson (1986) found that asset returns are indeed an increasing function of the spread suggesting that liquidity affect the asset returns. Several other papers have then validated this relation. Brennan & Subrahmanyam (1996) looked at the relationship between monthly stock returns and measures of illiquidity. By performing a cross-sectional analysis of the stock returns as a function of the cost of liquidity, they found a significant positive relation between the two. Other studies by Brennan et al. (1998) looked at the relationship between return and liquidity, by modeling liquidity as stock trading volume; they found that higher stock trading volumes was correlated with lower expected stock returns. Datar et al. (1998) also took on a similar approach and modeled liquidity with stock turnover, they also found a similar result, and higher stock turnover meant lower returns. Studies that are more recent have also looked at the relationship between liquidity, liquidity risk and asset returns, Acharya and Pedersen (2005), Pastor and Stambaugh (2003) has all suggested that the level of liquidity and liquidity risk is a priced risk factor. Bradrania & Peat (2014) also look at this topic by using a two-factor asset-pricing framework, and found that liquidity has an impact on expected return and that it is a risk factor to include. This statistically significant relation between expected return and liquidity would then have an important effect for the value of a firm's stock. If there is a premium for liquid stocks this means that the stock price would be higher compared to an illiquid stock. From the reasoning above, this means that an investor would demand a higher return for an illiquid stock, which is achieved by a lower stock price, everything else being equal (Amihud & Mendelson, 2000). Therefore, 13

20 there seems to be incentives for firms to try to enhance the liquidity of their stock and as a result get a higher stock price. B.1 Estimation methods A vast literature examines the evidence of liquidity premium across different asset classes using different methods. An overview of the different methods is presented below structured by asset type based on Hibbert et al. (2009) classification. For corporate bonds there are four approaches used: the microstructure approach, the direct approach, the structural model approach and the regression based approach. The microstructure approach does not provide a model for empirical testing however emphasizes on a framework describing how trading costs and market costs explain the different expected returns and yields. The direct approach involves a more straightforward method wherein two assets or asset portfolios have the same characteristics but with varying levels of liquidity that allows calculating the difference in yields. Brooke et al. (2000) apply this method on the collateral repo rates resulting in an estimate of liquidity premium at 15 bp. The third approach consists on comparing an estimated fair credit spread to actual market spreads to derive the liquidity premium. This method is similar to the direct approach in the sense that it compares an illiquid asset to its liquid counterpart. Webber (2007) provides an example of this approach by proposing a model built on Leland and Toft s (1996) method to derive default and demonstrates a rise in liquidity premium following a recent financial crisis. The regression analysis based approach has been widely used among researchers for two main purposes, one to identify the liquidity proxies that can be linked to higher yields, and second to attempt to quantify the liquidity premium by establishing the relationship between the liquidity proxy and the excess asset returns or yields. Han and Zhou (2007) and De Jong and Driessen (2012) conduct a regression analysis on corporate bonds. As an overview of the empirical evidence on liquidity premium of corporate bonds is that it is existent, substantial, varying over time. 14

21 For equity the main method used is regression analysis. Many studies advanced the literature on stock liquidity premium using a regression analysis after Amihud and Mendelson (1986). The main difference consisted on the different liquidity measures used in their research. Brennan and Subrahmanyam (1996) used the daily intraday trade and quote data to estimate liquidity costs. Loderer and Roth (2005) used relative bid-ask spreads as liquidity measure instead. Brennan et al. (1998) and Pástor and Stambaugh (2003) conducted their studies using stock trading volume as a measure of liquidity. On the other hand Datar et al. (1998) and Nguyen et al. (2007) use stock turnover as proxy for liquidity. Some studies focus not on proving the existence of a liquidity premium but instead aim to quantify it. Acharya and Pedersen (2005) develop a liquidity adjusted capital asset pricing model using Amihud (2002) measure for liquidity and estimate the return premium associated with liquidity at 3.5%. Bekaert et al. (2007) study the liquidity risk in emerging markets and estimated a local liquidity premium around 85 basis points per month. Lesmond (2005) studied the equity liquidity premium across 23 emerging markets and found differences in liquidity premium based on political and legal systems. Silva and Chávez (2008) analyses the liquidity premium in the Latin American equity markets. His findings showed an average of 8 % in liquidity premium across the four markets in the study. 15

22 C. Liquidity premium and the link to corporate finance Amihud and Mendelson hypothesize in their 1986 paper about a link between corporate finance policies and liquidity. They suggested, based on the evidence that more liquid stock required lower returns, that firms have an incentive to invest in liquidity-increasing financial policies in order to reduce the opportunity cost of capital thereby increasing the overall value of the firm. This implication on capital structure and firm value stems from the violation of one the assumptions in the Miller-Modigliani capital structure (Amihud and Mendelson, 2012). With the presence of high enough liquidity costs, like information and transaction costs, the firm s equity cost of capital will be increased. Furthermore they have suggested (Amihud and Mendelson 1988, 1991) four main firm liquidity-enhancing strategies: (a) To increase the company s investor base. Empirical evidence from a study of Japanese stock (Amihud et al. 1999) showed that improving the investor base resulted in increased stock liquidity and stock price. (b) To increase the information disclosure through direct and indirect dissemination. This strategy reduces the information asymmetries about firm value and thus increases stock price. (c) To avoid stock fragmentation. This strategy s concept arises from the fact that having multiple stock types reduces the stock s liquidity, therefore reducing individual stock price and overall firm value. Amihud et al. (2003) found that when two different set of assets from the same firm were merged, the stock became more liquid and rose in price. (d) To list on liquid exchange markets. Total liquidity contributes to the liquidity of the individual stock. 16

23 Further literature exists on other implications of liquidity on corporate finance. Bharath, Pasquariello and Wu (2009) explore the relationship between capital structure and stock liquidity. Their paper demonstrates that firms that have a higher debt level will have assets with lower liquidity. This illiquidity is explained by the exogenous information asymmetries that arise between the creditors who have more access to inside information and the shareholders. Furthermore they propose an incremental effect between information asymmetries and leverage. The higher the information asymmetries then the higher the likelihood to raise funds through debt, which will in turn cause higher information asymmetries. Additional literature by O Connor, Lesmond and Senbet (2008) also reinforces and demonstrates that the liquidity of a firm s stock will decrease as the level of debt increases. Their study consists on analyzing firms that had capital structure changes in their sample period in order to examine the impact that it has on stock liquidity measured by the bid ask spread and a model to measure the price impact of liquidity. Lipson and Mortal (2009) also provide evidence and conclude that liquidity has a significant impact on capital structure decisions. Eckbo and Norli (2000) find that in the years subsequent to the IPO, IPO stocks have significantly lower leverage ratios and higher liquidity (turnover) than control firms matched on size. In overall, studies (e.g. O Connor, Lesmond and Senbet, 2008; Bharath, Pasquariello and Wu, 2009) indicate that firms should take liquidity into account when managing the target level of debt of a firm. There is a tradeoff between the benefits of debt financing such as tax savings and the costs it causes from illiquidity. This effect can be further extended to the increase in the equity cost of capital. Another study by Banerjee, Gatchev and Spindt (2007) explores the liquidity hypothesis of dividends. The hypothesis of dividends states that in a market with trading frictions, other things equal, firms with less liquid stock are more likely to pay out dividends. They do indeed find proof backing up their hypothesis. Brockman, Howe and Mortal (2008) conducted similar research on liquidity and dividend policy. However in their study they set out to prove the liquidity hypothesis of repurchases which claims that stock liquidity influences the decisions on corporate payout policy primarily through repurchases instead of dividends. The results 17

24 showed that firms who initiated repurchases are generally more liquid than firms who did not, comparably firms who initiated dividend payouts were less liquid than those who did not. Their study therefore reinforces the idea that dividend payouts are a form of compensation for less liquid stock. Griffin (2010) adds to the literature on dividend policy and stock liquidity. Their study supports the relationship previously found by Banerjee, Gatchev and Spindt (2007). Moreover, Sarin, Shastri and Shastri (1999) show that liquidity decreases as the level of inside ownership increases. They empirically prove as well that the stock liquidity is inversely related to institutional ownership. This entails that firms have the ability to improve their stock liquidity by targeting an appropriate level of inside ownership and institutional ownership. Note that firms may only aim for this appropriate level of inside ownership and/or institutional ownership since they cannot directly control whether institutions purchase their shares or not. As explanation for this inverse relationship they suggest that the information asymmetries are an increasing function of insider holdings. In contrast to their study, a study by Glosten and Harris (1988) found an inconclusive and insignificant relation between stock liquidity and insider holdings using a sample of NYSE stocks from 1981 to Researchers have opposing results in regards to the relationship between stock liquidity and the percentage of insider holding. More recently Agarwal (2009) explores the relationship between liquidity and institutional ownership. He finds that an increase in institutional ownership may result in opposing effects on liquidity. On one hand it may decrease liquidity by increasing information asymmetry which he describes as an adverse selection effect. And on the other hand it may increase liquidity by increasing price discovery due to competition among institutions information efficiency effect. The relationship is described as non-monotonic (increasing liquidity at first until reaching optimality and then decreasing as institutional ownership increases). The liquidity of the stock also affects certain aspects related to corporate governance. Maug (1998) shows how liquid stock markets allow large investors to benefit from monitoring and it also helps to reduce the free-rider problem, thus improving corporate governance. Bolton and Von Thadden (1998) explore the costs and benefits of liquidity reflected by the firm s 18

25 ownership structure. They compare the benefits of liquidity obtained from a dispersed ownership to the benefits of efficient management stemmed from a degree of ownership concentration, thus lower liquidity. D. Firm Liquidity After exploring the literature of liquidity in general first, followed by liquidity in asset pricing and its link to corporate financial policies, we proceed with another aspect of liquidity in order to get the broad picture of all the different liquidity terms that will be dealt with throughout this thesis. Firm liquidity can be defined as the availability of internal funds (Hoshi, Kashyap, Scharfstein, 1991). In other words it can be described as the firm s ability to convert assets into cash in order to meet its financial obligations. The most liquid form of an asset is namely cash. Emery and Cogger (1982) state that firms maintain liquidity in order to meet near short term and long term expected and unexpected outflows. Among the widely used spread measures of firm liquidity are the current ratio which is current assets over current liabilities, the quick ratio which is the cash and cash equivalents over current liabilities, and net working capital which is current assets minus current liabilities (Bierman, 1960). Certain financial policies affect directly or indirectly the firm s liquidity. Increasing the dividend payout for example will decrease the firm s cash holdings thus the firm liquidity. Similarly, the corporate decision of setting a low cash holding target will have an impact of the firm s liquidity. Gopalan, Kadan and Pevzner (2012) found a positive relationship between stock liquidity and the liquidity of the assets of a company (what we described as firm liquidity). Their study showed how asset liquidity improved stock liquidity more for firms which are less likely to reinvest their liquid assets. Their findings also revealed that an increase in corporate cash holdings has higher benefits for firms with more illiquid stock. 19

26 E. Earlier research in Sweden The relationship between stock returns and liquidity is not a well-researched topic when it comes to the Swedish stock market. There have just been a few previous studies investigating the concept of liquidity and the existence of a liquidity premium in Sweden. The findings of these studies are somewhat ambiguous in the support of the existence of a liquidity premium. Both Pavlica and Persson (2012), Svartholm and Uhrberg (2012), and Gerwin (2005) tried to validate the findings of Amihud and Mendelson (1986) by looking at the relationship between stock return and liquidity on the Swedish stock market. All of the studies took on a similar approach to try to investigate the relationship; they did portfolio analysis and cross-sectional regressions based on time series data. The data consisted of time series data of a variety of companies together with other variables such as the bid-ask spread as a proxy for measuring liquidity. Gerwin found only weak evidence regarding the relationship between liquidity and stock returns. Pavlica and Persson (2012) also did not find clear evidence for the existence of a liquidity premium. However, the studies argues that their sample periods can help to explain the ambiguous results they got with the later having a time period that is very volatile. Other studies such as Personne (2013) also investigates the same relationship to see if liquidity is a factor influencing returns on the Swedish stock market for the period The author fits a Sharpe-Lintner CAPM model with an illiquidity factor to series of excess return. He finds that illiquidity has an impact on returns. 20

27 III. Methodology A. Sample and Data Our data sample consist of Swedish companies listed on both OMX and Aktietorget, we first collected the available data for all Swedish companies from Reuters DataStream. This gave us a sample size of approximately 900 companies. For each of these companies our desired explanatory variables where retrieved on a monthly basis for the time period to The data needed to calculate our liquidity proxies where retrieved on a daily basis for the same time period to then calculate monthly averages to make the data consistent. The time period were selected based on two main reasons, first one being that we wanted as long of a period as possible to even out the volatility associated with the financial crisis of Second reason for not having a longer period of time is that there were limited data available before We then started to reduce companies in a systematic way based on our liquidity variables. Our criterion was that companies that had no available information on our selected independent variables were to be deleted. This left us with a sample size of 433 companies, with 180 monthly observations each giving us a total of 77,940 observations. The data was structured as panel data, with companies as the cross sectional dimension. Several limitations were faced when conducting this research study. One limitation was the fact that the data was unbalanced. On a multivariate regression Eviews removes every period point that does not have a value for all the variables, thus reducing furthermore the sample size. In our sample we had a large number of observations as previously mentioned, nevertheless when running the regressions the sample size is reduced to approximately 27,000. Furthermore Eviews has certain limited functions regarding panel data validity tests, driving us to conduct manual tests to control for aspects such as heterogeneity for example. 21

28 B. Liquidity Variables We will use the aforementioned measurements: Bid-ask spread, Amihud-measure and the Turnover to proxy liquidity. The reason for using three measurements is to investigate if our firm level variables are significant across the suggested liquidity measurements. We have described the main logic behind these proxies and their formulas in the previous chapter. In this section, we will go further into why the particular measurements have been chosen and how we will calculate them based on our data sample. In order to see how the liquidity variables have evolved through our sampled time period we graphed the yearly average of ten randomly selected companies for each measurement as can be seen in Appendix A. The Bid-ask spread (BA_S) is used because of its characteristics but also because it is easily accessible in Thomson Reuter s DataStream. Furthermore, we use the bid-ask spread because of previous studies showing that it is a valid proxy for liquidity. Amihud and Mendelson (1986), Brennan and Subrahmanyam (1996), Amihud and Mendelson (2005) and Dimson and Henke (2002) all use the bid-ask spread in their studies to model liquidity. We will use the quoted bid and ask prices of the companies in our data sample on a daily basis. In DataStream these daily quotes is defined as the latest bid and ask price. The spread itself is then calculated as the Ask price minus the Bid price divided by its midpoint, we will then take the average of these daily bid-ask spreads to obtain it on a monthly basis. In the graph depicting the evolution of the bidask spread in Appendix A we observe two spikes in the spread indicating a decrease in liquidity. These events correspond to the internet bubble in early 2000 and the financial crisis of The Amihud-measure is as previously mentioned another popular proxy for liquidity because it captures the illiquidity of an asset. This measurement has also been used by a variety of previous studies such as Goyenko, Holden and Trzcinka (2009), Acharya and Pedersen (2005) and Dick- Nielsen et al (2012) to proxy liquidity. The Amihud-measure is also popular because of the 22

29 accessibility of the data on which it is calculated. We will use the quoted daily stock prices to calculate the daily stock returns on the companies in our data sample. To get the Amihudmeasure on a daily basis we will then divide the daily stock returns by the daily SEK volume traded. We will then take the average on these daily Amihud-measures to get a monthly Amihud measurement: ILLIQ. The graph of the Amihud measure evolution over our sample period (Appendix A) shows two main decreases in liquidity (increases in the Amihud measure) corresponding to 2008 and These observations may be the result of the 2008 financial crisis as well as the euro crisis of This measurement does not seem to capture any effect from the internet bubble in 2000 as does the bid-ask spread. Finally, we use the Turnover as our third proxy for liquidity. A number of studies has suggested that the turnover can be used to proxy liquidity sense it indicates how transacted a specific asset is. The turnover (TURN_VAL) is defined as the total trading volume of an asset over the number of outstanding assets. This measurement has also been used in previous studies such as Brooke et al. (2000) to model liquidity. We have obtained the turnover for the companies in our data sample from Thomson Reuter s DataStream on a daily basis; this is defined as the number of shares traded on a particular day. To get the turnover ratio we will divide this number by the number of outstanding shares. We will then use the average of these to get it on a monthly basis. In appendix A the time trend of turnover is also graphed. We observe two increases in liquidity in the years 2001 and These years correspond to periods under stress nevertheless there seems to be an increase in liquidity based on this measurement. The reason for this could be due to the noisiness included in the measurement, turnover basically captures different aspects of trading and not just liquidity. 23

30 C. Explanatory Variables and Control Variables The aim of our study is to identify financial policies, which have a significant relation with the firm s stock liquidity given the positive relation between stock liquidity and firm value. Based on previous literature discussed in chapter II we have selected a set of firm adjustable factors, (referred to as explanatory variables) as well as several control variables (which will be denoted as non-firm adjustable) to run a regression analysis. By firm adjustable factors we imply a set of internal factors that a firm is able to actively modify through different financial policies. On the other hand, the non-firm adjustable ones are external factors which, even though they are individually related to the firm, are difficult to change by management. Each factor will be measured using one or more variables. The intuition and measures used will be detailed next. C.1 Explanatory Variables Dividend Policy: The dividend policy has been shown to affect the liquidity of the stock (Banerjee, Gatchev and Spindt, 2007; and Griffin, 2010). Additionally firms are able to actively increment, maintain, lower or even cancel the dividend payouts in a firm. For the purpose of this study the fact that managers are normally reluctant to cut down dividends or initiate a dividend payment is ignored. This normally occurs since this cutting down dividends may suggest to investors that they feel the company s prospects are not good enough to support a dividend payment any longer (Black, 1976). Similarly, initiating a dividend payment will require them to maintain it for a period of time. The variable used to measure dividend policy is the firm s dividend yield (DIV_Y) calculated as the dividend per share as a percentage of the share price per month. Firm Liquidity: 24